The main problem with legs is there is not a good analysis for them. FEA is one solution. You might consider if you can treat the leg attachment as an upside-down anchor chair per standard anchor chair methods. Or you can design a circular girder so that the girder itself carries all non-axisymmetric loads, leaving the tank itself axisymmetrically loaded. In many cases, on small tanks, it is more economical to beef up the connection details than to spend engineering time designing better ones.
Assuming you work out some way to adequately analyze the tank, there's no reason you couldn't put the legs on the elliptical head. I've noticed a bunch of little cone-bottom tanks on farms, and most of them have legs attaching to the cone with circular ring.